Publications

Diabetic polyneuropathy is a major complication of diabetes mellitus, causing severe alterations of the neural circuits between spinal nerves and spinal cord. The analysis of 3D confocal images of dorsal root ganglia in diabetic mice, where different fluorescent markers are used to identify different types of nociceptors, can help understanding the unknown mechanisms of this pathology. Nevertheless, due to the inherent challenges of 3D confocal imaging, a thorough and comprehensive visual investigation is very difficult. In this work we introduce a tool, 3DRG, that provides a fully-automated segmentation and 3D rendering of positively labeled nociceptors in a dorsal root ganglion, as well a quantitative characterisation of its immunopositivity to each fluorescent marker. Our preliminary experiments on 3D confocal images of entire dorsal root ganglia from healthy and diabetic mice provided very interesting insights about the effects of the pathology on two different types of nociceptors.

HMMs are widely used in action and gesture recognition due to their implementation simplicity, low computational requirement, scalability and high parallelism. They have worth performance even with a limited training set. All these characteristics are hard to find together in other even more accurate methods. In this paper, we propose a novel doublestage classification approach, based on Multiple Stream Discrete Hidden Markov Models (MSD-HMM) and 3D skeleton joint data, able to reach high performances maintaining all advantages listed above. The approach allows both to quickly classify presegmented gestures (offline classification), and to perform temporal segmentation on streams of gestures (online classification) faster than real time. We test our system on three public datasets, MSRAction3D, UTKinect-Action and MSRDailyAction, and on a new dataset, Kinteract Dataset, explicitly created for Human Computer Interaction (HCI). We obtain state of the art performances on all of them.

This paper presents a novel retrieval pipeline for video collections, which aims to retrieve the most significant parts of an edited video for a given query, and represent them with thumbnails which are at the same time semantically meaningful and aesthetically remarkable. Videos are first segmented into coherent and story-telling scenes, then a retrieval algorithm based on deep learning is proposed to retrieve the most significant scenes for a textual query. A ranking strategy based on deep features is finally used to tackle the problem of visualizing the best thumbnail. Qualitative and quantitative experiments are conducted on a collection of edited videos to demonstrate the effectiveness of our approach.

Optical coherence tomography (OCT) represents a non-invasive imaging technology, which may be applied to the diagnosis of non-melanoma skin cancer and which has recently been shown to improve the diagnostic accuracy of basal cell carcinoma. Technical developments of OCT continue to expand the applicability of OCT for different neoplastic and inflammatory skin diseases. Of these, dynamic OCT (D-OCT) based on speckle variance OCT is of special interest as it allows the in vivo evaluation of blood vessels and their distribution within specific lesions, providing additional functional information and consequently greater density of data. In an effort to assess the potential of D-OCT for future scientific and clinical studies, we have therefore reviewed the literature and preliminary unpublished data on the visualization of the microvasculature using D-OCT. Information on D-OCT in skin cancers including melanoma, as well as in a variety of other skin diseases, is presented in an atlas. Possible diagnostic features are suggested, although these require additional validation.

State of the art approaches for saliency prediction are based on Full Convolutional Networks, in which saliency maps are built using the last layer. In contrast, we here present a novel model that predicts saliency maps exploiting a non-linear combination of features coming from different layers of the network. We also present a new loss function to deal with the imbalance issue on saliency masks. Extensive results on three public datasets demonstrate the robustness of our solution. Our model outperforms the state of the art on SALICON, which is the largest and unconstrained dataset available, and obtains competitive results on MIT300 and CAT2000 benchmarks.

This paper discusses the role of computer vision to bridge the experiential gap between the cultural and emotional experience of the visitors in museums or cultural heritage sites. We don't argue against the use of multiple sensors to provide a more complete cultural experience but claim the primary role of computer vision for such a task. Although many research challenges are still far to be solved effectively, especially for detection, re-identification, tracking and recognition, we believe that technology can be deployed already in real contexts and support concrete applications with interesting results that will open the door to valuable future applications.